Why are rural hospitals closing in the U.S.? Predictors identified using explainable machine learning.

Area of Science:

• Healthcare Management
• Health Services Research
• Machine Learning in Healthcare

Background:

• Rural hospital closures exacerbate healthcare access issues and negatively impact community health and economic outcomes.
• Identifying risk factors for rural hospital closures is crucial for developing targeted interventions.

Purpose of the Study:

• To identify and interpret the key risk factors associated with rural hospital closures in the U.S.
• To leverage explainable machine learning (XML) and national longitudinal data for this analysis.

Main Methods:

• A retrospective longitudinal study of U.S. rural hospitals (2011-2022) using integrated financial, operational, policy, population, and community data.
• Utilized XGBoost classifier with cross-validation and hyperparameter tuning, employing SHapley Additive exPlanations (SHAP) for model interpretability.
• Addressed data challenges including class imbalance and multicollinearity.

Main Results:

• The predictive model achieved 86% recall, 75% accuracy, and 89% AUC.
• Key closure risk factors include sustained financial and operational instability, policy context (e.g., Medicaid expansion timing), and community sociodemographics.
• Community factors may reflect underlying socioeconomic conditions influencing closure risk.

Conclusions:

• Rural hospital closure risk is a result of accumulated long-term financial and operational instability, influenced by policy and community contexts.
• Longitudinal data and XML offer a practical framework for early warning systems to support proactive monitoring and interventions.
• This approach is particularly valuable for hospitals serving historically underserved populations.